Date of Award

Summer 2019

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

Committee Director

Danella Zhao

Committee Member

Li Yaohang

Committee Member

Sampath Jayarathna

Abstract

Electrical grid maintenance and repairs are crucial services that keep America’s lights on. Electrical service providers make it their priority to uphold minimal interruptions to this service. Electricity is essential for modern technology within the home, such as cooking, refrigeration, and hot water. Organizations, such as schools, hospitals, and military bases, cannot properly function or operate without power. When analyzing the current electrical infrastructure, it is evident that considerable components of the power grid are aging and in need of replacement. Additionally, threats and damage continue to occur. These damages occur not only due to simple, single power line failure but also on a larger scale in the event of natural disasters. Instead of replacing current aging components or sending out crews of people for preventative maintenance and repairs, neural networks provide innovative technology that can improve these processes. With the use of unmanned aerial vehicles (UAVs), neural networks can identify and classify both normal functioning and damaged electrical power lines.

This thesis will investigate the use of convolutional neural networks and low-cost unmanned aerial vehicles (UAV)’s to identify and detect damage to power lines that carry electrical service to consumers called distribution lines. The UAVs can serve as a vehicle to supply neural networks with input imagery data and automatically evaluate the condition of power lines. These neural networks are comprised of many layers that have been configured for this specific use and provide efficient identification and detection performance. Together, the UAV-neural network system can provide more efficient routine maintenance with wider coverage of areas, increased accessibility, and decreased time between identification of issues and subsequent repair. Most importantly, the use of neural networks will keep electrical crews safe and provide faster response in the setting of natural disaster. In this day and age, we must think smarter and respond more efficiently to serve continually growing areas and reach areas with less resources.

DOI

10.25777/f1za-0j73

Share

COinS